This is a repository copy of A common intronic variant of PARP1 confers melanoma risk and mediates melanocyte growth via regulation of MITF.

White Rose Research Online URL for this paper:http://eprints.whiterose.ac.uk/119196/

Version: Accepted Version

Article:

Choi, J, Xu, M, Makowski, MM et al. (16 more authors) (2017) A common intronic variant ofPARP1 confers melanoma risk and mediates melanocyte growth via regulation of MITF. Nature Genetics, 49 (9). pp. 1287-1413. ISSN 1061-4036

https://doi.org/10.1038/ng.3927

© 2017 Nature America, Inc., part of Springer Nature. This is an author produced version of a paper published in Nature Genetics. Uploaded in accordance with the publisher's self-archiving policy.

eprints@whiterose.ac.ukhttps://eprints.whiterose.ac.uk/

Reuse

Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy solely for the purpose of non-commercial research or private study within the limits of fair dealing. The publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White Rose Research Online record for this item. Where records identify the publisher as the copyright holder, users can verify any specific terms of use on the publisher’s website.

Takedown

If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing eprints@whiterose.ac.uk including the URL of the record and the reason for the withdrawal request.

A common intronic variant of PARP1 confers melanoma risk and mediates melanocyte 1

growth via regulation of MITF 2

Jiyeon Choi1,7, Mai Xu1,7, Matthew M. Makowski2, Tongwu Zhang1, Matthew H. Law3, Michael A. 3

Kovacs1, Anton Granzhan4, Wendy J. Kim1, Hemang Parikh1, Michael Gartside3, Jeffrey M. 4

Trent5, Marie-Paule Teulade-Fichou4, Mark M. Iles6, Julia A. Newton-Bishop6, D. Timothy 5

Bishop6, Stuart MacGregor3, Nicholas K. Hayward3, Michiel Vermeulen2, Kevin M. Brown1* 6

1Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD; 7

2Department of Molecular Biology, Faculty of Science, Radboud Institute for Molecular Life 8

Sciences, Radboud University Nijmegen, the Netherlands; 3QIMR Berghofer Medical Research 9

Institute, Brisbane, QLD, Australia; 4CNRS UMR9187, INSERM U1196, Institut Curie, PSL 10

Research University, and Université Paris Sud, Université Paris Saclay, Orsay, France; 11

5Translational Genomics Research Institute, Phoenix, AZ; 6Section of Epidemiology and 12

Biostatistics, Leeds Institute of Cancer and Pathology, University of Leeds, Leeds, UK; 7These 13

authors contributed equally to this work. *Correspondence should be addressed to K.M.B at: 14

Email: kevin.brown3@nih.gov 15

ABSTRACT 16

Prior genome-wide association studies have identified a melanoma-associated locus on 17

chr1q42.1 that encompasses a ~100 kb region spanning the PARP1 gene. eQTL analysis in 18

multiple cell types of melanocytic lineage consistently demonstrated that the 1q42.1 melanoma 19

risk allele (rs3219090, G) is correlated with higher PARP1 levels. In silico fine-mapping and 20

functional validation identified a common intronic indel, rs144361550 (-/GGGCCC, r2 =0.947 21

with rs3219090) as displaying allele-specific transcriptional activity. A proteomic screen 22

identified RECQL as binding to rs144361550 in an allele-preferential manner. In human primary 23

melanocytes, PARP1 promotes cell proliferation and rescues BRAFV600E-induced senescence 24

phenotypes in a PARylation-independent manner. PARP1 also transforms TERT-immortalized 25

melanocytes expressing BRAFV600E. PARP1-mediated senescence rescue is accompanied by 26

transcriptional activation of melanocyte lineage survival oncogene, MITF, highlighting a new role 27

of PARP1 in melanomagenesis. 28

29

To date, genome-wide association studies (GWAS) have identified twenty common, 30

genome-wide significant melanoma susceptibility loci1-9, most of which do not appear to be 31

explained by protein-coding variants. A subset of these loci harbor known pigmentation genes 32

that mediate melanoma-associated phenotypes such as eye, hair, and skin color. While several 33

loci harbor genes implicated in cancer, evidence directly linking common risk variants within 34

most of these loci to altered function of specific genes is lacking. 35

MacGregor and colleagues initially identified a melanoma risk locus tagged by 36

rs3219090 on chromosome band 1q42.1 in an Australian case-control study at a near genome-37

wide level of significance (P = 9.3 x 10-8 , OR = 0.87, protective allele A)8. The association has 38

since been replicated by multiple other studies3,10, including most recently by a meta-analysis of 39

12,874 melanoma cases (rs1858550, P = 1.7 x 10-13)7. Notably, the locus at 1q42.1 has also 40

been associated with melanoma survival11, where the melanoma risk allele correlates with 41

increased survival, an association that has since been replicated12. The region of association 42

spans from 226.52 Mb to 226.63 Mb (hg19) of chromosome 1, encompassing the entirety of the 43

poly(ADP-ribose) (PAR) polymerase-1 (PARP1) (OMIM: 173870) gene, and fine-mapping 44

suggests that the association is best explained by a single-SNP model3. 45

While a number of other genes are located in the vicinity of the association peak, PARP1 46

has the most well-established role in cancer. PARP1 is best known for its role as a DNA repair 47

enzyme and genotoxic sensor that functions in base excision repair (BER), single-strand break 48

repair, and double-strand break repair13. Once PARP1 binds to damaged DNA, its enzymatic 49

function is activated, and it covalently attaches PAR polymers to acceptor proteins, including 50

histones and PARP1 itself14. PARP1 amplifies DNA damage signals, modifies chromatin 51

structures to accommodate DNA damage response proteins, and further recruits DNA repair 52

proteins13,15,16. While PARP1 is not directly involved in repair of UV signature mutations via 53

nucleotide excision repair, its role in the repair of DNA lesions induced by reactive oxygen 54

species (ROS) is well-established17. ROS are generated by UVA exposure18, are a byproduct of 55

melanin production19, and appear to play a role in oncogene-induced senescence (OIS)20,21. 56

Aside from DNA repair, PARP1 functions in regulating gene expression by modifying chromatin 57

structure, associating with promoters and enhancers, and acting as a transcriptional co-58

regulator22,23. While many of these roles rely on PARP1 catalytic activity, some are also 59

PARylation-independent, as in the transcriptional co-regulator function for NF-kB and B-60

MYB24,25. 61

In this study, we functionally characterized the 1q41.2 melanoma risk locus, 62

demonstrating a consistent correlation of the risk genotype with levels of PARP1 gene 63

expression in tissues of melanocytic origin, identifying a gene regulatory variant within the first 64

intron of PARP1, and elucidating a role for PARP1 in melanocyte OIS via regulation of the 65

melanocyte master regulatory transcription factor, MITF. 66

67

RESULT 68

The rs3219090 risk allele is correlated with high PARP1 69

We performed expression quantitative trait locus (eQTL) analysis in order to identify 70

genes for which expression levels are correlated with 1q42.1 risk genotype in tissues of 71

melanocytic lineage. Initially we evaluated the correlation of rs3219090 with expression of 72

genes within +/-1Mb in 59 early-passage melanoma cell lines using expression microarray data. 73

The results indicated that the rs3219090 risk allele is associated with higher levels of PARP1 74

expression (P = 1.4 x 10-3, linear regression; Fig. 1a). Notably, PARP1 is the only gene in the 75

region that passed a Bonferroni-corrected P-value threshold (corrected for 14 genes, P < 3.6 x 76

10-3; Supplementary Table 1), and this eQTL subsequently validated via qPCR assay (P = 77

0.031, linear regression; Supplementary Fig. 1a). We then sought independent replication of 78

PARP1 and other nominally significant eQTL genes (P < 0.05) in publicly available RNA-79

sequencing datasets for melanoma-relevant tissues. When 409 melanoma tumors from The 80

Cancer Genome Atlas (TCGA) project (dbGAP Accession: phs000178.v9.p8) were tested, the 81

melanoma risk allele of rs3219090 was again significantly correlated with higher PARP1 82

expression levels (P = 3.9 x 10-3, linear regression using copy number as a covariate; 83

Supplementary Fig. 1b) while no other genes were significantly correlated (Supplementary 84

Table 2). Similarly, the PARP1 eQTL was replicated in normal skin samples collected through 85

the Genotype-Tissue Expression (GTEx) Project (dbGAP Accession: phs000424.v6.p1), 86

including those derived from both sun-exposed skin (P = 2 x 10-4, linear regression, n = 302) 87

and non-sun-exposed skin (P = 0.011, linear regression, n = 196) (Supplementary Fig 1c-d, 88

Supplementary Table 3-4). Together, these data identified PARP1 as the strongest eQTL gene 89

in the chr1q42.1 locus whose expression displayed the most consistent correlation with 90

genotypes of the lead SNP in sample panels of melanocytic lineage as well as human skin. 91

To complement eQTL data and rule out the possibility of any sample-specific 92

confounding factors masking genotype effect, we performed allele-specific expression (ASE) 93

analysis for PARP1 in samples carrying both risk and protective alleles. Fourteen melanoma cell 94

lines that are heterozygous for rs3219090 and harbor normal regional copy number were 95

assayed using a quantitative allelic TaqMan assay for a synonymous coding surrogate SNP 96

(rs1805414; r2 = 0.98 with rs3219090 in 1KG phase3 EUR), where allelic ratio was inferred from 97

known ratios of allelic standards. The results demonstrated a significant allelic imbalance 98

towards a higher proportion of PARP1 expressed from the risk allele in the majority of 99

heterozygous cell lines (P = 1.2 x 10-4, two-tailed Wilcoxon signed rank test; Fig. 1b). Significant 100

allelic imbalance was also observed when a subset of these cell lines were analyzed by 101

RNAseq (data not shown). Subsequent PARP1 ASE analysis in TCGA and GTEx RNAseq 102

datasets demonstrated that a higher allelic proportion of mapped reads was also observed for 103

the risk allele across TCGA tumor samples (P = 0.011, two-tailed Wilcoxon signed rank test, n = 104

48, copy-neutral and heterozygous; Fig. 1c), as well as in sun-exposed and non-sun-exposed 105

skin tissues (GTEx, P = 1.16 x10-5, n = 139; P = 8.9 x 10-5, n = 69; respectively, two-tailed 106

Wilcoxon signed rank test; Supplementary Fig. 1e-f). These data demonstrate that the 107

melanoma risk allele of rs3219090 is significantly associated with increased PARP1 expression 108

in tissues of melanocytic origin and skin with striking consistency across multiple datasets. 109

110

Fine-mapping and functional annotation of candidate SNPs 111

Given that high PARP1 levels are correlated with the melanoma risk allele of rs3219090, 112

we next sought to identify functional risk variant(s) that may influence PARP1 expression. 113

Previously, fine-mapping of this locus in a large European population provided support for a 114

model in which a single variant accounts for the association signal in this region3, a finding 115

confirmed as part of the meta-analysis conducted by Law and colleagues7. We prioritized 65 116

variants that are highly correlated with the lead SNP as candidate functional variants (r2>0.6 117

with lead SNPs from the discovery or meta-analysis lead SNPs3,7, rs3219090 and rs1858550, 118

respectively; LD based on 1KG phase3, EUR and CEU). Given the absence of amino acid-119

changing PARP1 variants within this set of candidates, an absence of evidence for alternative 120

splicing as a likely mechanism (Supplementary Note), and considerable evidence for allelic 121

differences in PARP1 expression levels, we focused on those located within annotated 122

melanocyte- or melanoma-specific cis-regulatory elements using data from the ENCODE26 and 123

Roadmap projects27 (Supplementary Note, Supplementary Table 5-6, Supplementary Fig. 124

2-3). The four of the most strongly supported variants are situated at the center of melanocyte 125

DHS peaks as well as within regions harboring promoter or enhancer histone marks (H3K4me1, 126

H3K4me3, or H3K27ac) in the majority of melanocyte/melanoma cultures assayed 127

(Supplementary Table 6). Based on these data, we proceeded with functional characterization 128

of these four candidates (Supplementary Table 6, Supplementary Fig. 2). 129

130

An intronic indel displays allelic transcriptional activity 131

We assessed all four candidate functional variants for gene regulatory potential using 132

luciferase reporter assays, as well as for allelic patterns of protein binding via electrophoretic 133

mobility shift assay (EMSA). For these assays, we sought to identify variants that display 1) 134

transcriptional activation consistent with ENCODE annotation, 2) higher activity for the risk allele 135

consistent with the eQTL data, and 3) allele-specific protein binding. Among four candidate 136

variants, only rs144361550, a GGGCCC indel variant, met all these criteria (Fig. 2-3, 137

Supplementary Figs. 4-6; summarized in Supplementary Table 7). Namely, luciferase assays 138

conducted in a melanoma cell line demonstrated that the genomic region around rs144361550 139

exhibits strong transcriptional activity in both long (905bp covering the larger DHS region, ~17-140

20 fold higher than control levels) and short cloned fragments (22 or 28bp covering the 141

GGGCCC repeats, ~1.7-2.5 fold higher than control levels; Fig. 3a), where the risk-associated 142

deletion allele exhibited higher reporter activity than the insertion allele (30-45% higher). In 143

primary melanocytes, where transfection efficiency is considerably lower, allelic activity was not 144

observed, but the long deletion and insertion fragments displayed weak but significant 145

transcriptional activity (P = 1.2 x 10-3 and 5.9 x10-4, respectively, two-tailed, paired t-test; Fig. 146

3c). EMSAs using nuclear extract from melanoma cell lines or cultured primary human 147

melanocytes displayed preferential binding of nuclear proteins to the insertion allele (Fig. 3b,d). 148

Given the potential for miscalling genotype of this functional indel, we directly genotyped 149

rs144361550 in a large reference set to confirm LD with the lead SNP (Supplementary Note, 150

Supplementary Table 8-9, Supplementary Fig. 7-8). 151

To identify proteins that bind rs144361550 in an allele-preferential manner, we utilized 152

quantitative mass-spectrometry employing dimethyl label swapping28,29 30. Mass-spectrometry 153

using melanoma cell line extract identified exclusively insertion allele-preferential interactors, the 154

majority of which are not conventional transcription factors, including the RECQL helicase (Fig. 155

4a). While two transcription factors previously found by the ENCODE Project to localize to the 156

region overlapping rs144361550 via chromatin immunoprecipitation (ChIP) were found to bind 157

rs144361550 probes (TFAP2A, ZBTB7A), neither did so in an allele-preferential manner (data 158

not shown), in line with the observation that rs144361550 creates no new sequence motifs but 159

rather extends a poly-G repeat stretch. We then performed a series of antibody supershifts and 160

EMSAs using purified recombinant proteins for multiple candidates and validated that RECQL is 161

an unequivocal allele-preferential binder to rs144361550 (Fig. 4b, Supplementary Figs. 9, and 162

Supplementary Table 10). ChIP assays indicated that RECQL indeed binds to the PARP1 163

indel region in melanoma cells and primary human melanocytes carrying an insertion allele (Fig. 164

4c, Supplementary Fig. 10). We also performed a series of in silico and in vivo assays testing 165

for alternative DNA secondary structure formation (G-quadruplex or G4), with the results 166

suggesting RECQL-specific allelic binding mechanism rather than the one through G4 167

(Supplementary Note, Supplementary Table 10-11, Supplementary Fig. 11-13). 168

Ectopic expression of RECQL in three melanoma cell lines carrying insertion or deletion 169

alleles at a moderate level using lentiviral transduction resulted in a mild increase in PARP1 170

transcription (Fig. 4d). We then performed luciferase assays for rs144361550 with or without 171

RECQL over-expression in cells with low baseline levels of RECQL relative to melanomas 172

(HEK293FT cells). At a basal level, insertion and deletion alleles did not display differential 173

luciferase activity, but upon RECQL over-expression, significant allele-specific transcriptional 174

activity we previously observed in melanoma cell lines was recapitulated (Fig. 4e). Together, 175

these data suggest that RECQL may play a role in PARP1 allelic expression in cells of 176

melanocytic lineage through the melanoma risk-associated indel, rs144361550. 177

178

PARP1 facilitates melanocyte growth and transformation 179

Given that the risk allele of rs3219090 is associated with increased levels of PARP1 180

expression, we tested whether increased PARP1 levels could lead to altered cellular 181

phenotypes relevant to melanomagenesis. As BRAFV600E -induced senescence is well accepted 182

as a barrier to malignant transformation in early events of melanomagenesis 31-33 and ROS 183

seems to present a potential functional link between PARP1 and OIS 21, we examined whether 184

elevated PARP1 modulates OIS by overexpressing PARP1 in human primary melanocytes 185

expressing oncogenic BRAF (BRAFV600E). Consistent with published results, ectopic expression 186

of BRAFV600E in human melanocytes induced a robust arrest of cell growth and proliferation, 187

accompanied by heterochromatic H3K9Me3 focus formation (Fig. 5a-c, Supplementary 188

Fig.14), and a moderate increase in く-galactosidase activity (Supplementary Fig. 15), both 189

hallmarks of OIS. Overexpression of PARP1 prior to induction of BRAFV600E, however, 190

prevented the cell cycle arrest and H3K9Me3 focus formation as well as く-galactosidase activity 191

observed in melanocytes expressing BRAFV600E alone (Fig. 5a-c, Supplementary Fig. 14-15), 192

demonstrating a rescue from BRAFV600E-induced senescence. Crystal violet staining of cells 193

three weeks following PARP1 expression indicated that PARP1 by itself can also increase cell 194

proliferation in the absence of BRAFV600E, suggesting an effect on cell proliferation by PARP1 195

(Fig. 5a). PARP1 consistently exerted weak but significant effects on cell proliferation and OIS 196

reversal even at a moderate induction level (~1.5 fold, Supplementary Fig. 16) recapitulating 197

the subtle allelic expression differences observed in melanomas carrying risk or protective 198

alleles. Several oncogenes, including MYC 34 and PIK3CA 33, have been reported to stimulate 199

malignant transformation in melanoma cells by abrogating OIS and restarting cell proliferation. 200

To further evaluate if PARP1 can stimulate malignant transformation by affecting melanocyte 201

proliferation, we examined the effect of PARP1 on anchorage-independent growth of TERT-202

immortalized human melanocytes (p’mel35) by soft-agar assay. PARP1 cooperated with 203

BRAFV600E to enhance colony formation of p’mel cells in soft agar, similar to the previously 204

reported effect of MITF35, albeit to a lesser degree (Fig. 5d-e). Thus, like deregulated MITF 205

expression, increased expression of PARP1 can both rescue and further transform human 206

melanocytes from BRAFV600E-induced senescence. 207

Most of the well-characterized functions of PARP1, including those involving DNA repair, 208

are closely linked to its poly-ADP ribosylation activity. We therefore tested whether PARP1-209

mediated rescue of melanocytes from BRAFV600E-induced OIS is dependent on PARP1 catalytic 210

activity. Melanocytes overexpressing both PARP1 and BRAFV600E were treated with the PARP1 211

inhibitor BYK20416536, and the inhibition of PAR-activity was confirmed by Western blotting with 212

an antibody recognizing PAR (Fig. 6a). BYK204165 treatment did not block the cell proliferation 213

effect of PARP1 in BRAFV600E-expressing cells as shown by similarly increased cell growth in 214

crystal violet staining (Supplementary Fig. 17a), an increased percentage of BrdU-positive 215

cells and reduced G0/G1 population in PARP1-expressing cells (Fig. 6b), nor did it affect 216

H3K9Me3 focus formation (Supplementary Fig. 17b). We also tested a catalytically inactive 217

mutant version of PARP137-39 for effect on melanocyte growth and proliferation (Supplementary 218

Fig. 18a-d), resulting in similar phenotypes to those observed using BYK204165. 219

220

PARP1 transcriptionally activates MITF 221

To begin to understand the mechanism of PARP1-mediated senescence rescue we first 222

examined ROS content changes in BRAFV600E and/or PARP1 over-expressing melanocytes. 223

While we observed a minor increase of ROS content by BRAFV600E expression, we did not 224

detect meaningful changes in ROS content by PARP1 in our melanocyte system (data not 225

shown). We then turned our attention to melanocyte lineage survival oncogene, MITF, which is 226

frequently amplified in malignant melanomas and can transform melanocytes in the context of 227

BRAFV600E 35. Previous studies have observed that MITF expression is inhibited by oncogenic 228

BRAFV600E in melanomas40. Consistent with this finding, we demonstrate that persistent 229

activation of MAPK pathway by BRAFV600E also suppressed MITF expression in primary 230

melanocytes (Fig. 6a). Significantly, PARP1 induction, at either strong (Fig. 6a, Supplementary 231

Fig. 19) or more modest levels (Supplementary Fig. 16e-f), increases MITF at both mRNA and 232

protein levels, in a PARylation-independent manner (Supplementary Fig. 18d-f, Fig. 6a). 233

PARP1 also partially restores MITF expression in BRAFV600E-expressing melanocytes, 234

suggesting a link between OIS rescue and MITF restoration (Fig. 6a, Supplementary Fig. 235

16e,f, Supplementary Fig. 18d-f, Supplementary Fig. 19). Consistent with this finding, we 236

observed a weak but significant positive correlation between PARP1 and MITF transcript levels 237

(Supplementary Fig. 20) in 409 melanoma samples from TCGA (Pearson’s r = 0.198, P = 5.7 x 238

10-5) as well as in the smaller subset of 189 tumors that are copy-neutral at the MITF locus 239

(Pearson’s r = 0.252, P = 4.8 x 10-4). To address potential issues with tumor heterogeneity and 240

random factors driving this correlation in melanomas, we also looked at the correlation in our 241

early-passage melanoma cell lines. While we did not observe a significant correlation in the full 242

set of 59 cell lines (Pearson’s r = 0.15, P = 0.26; Supplementary Fig. 21a), when we further 243

subdivide them into MITF-high and MITF-low groups to account for distinct cellular states 244

signified by MITF levels41, a strong correlation of PARP1 and MITF levels was observed only in 245

MITF-high group (Pearson’s r = 0.560, P = 5.5 x 10-3, n = 23; P = 6.6 x 10-3 when MITF copy 246

number is adjusted; Supplementary Fig. 21b). A trend of association with PARP1 levels were 247

also observed in a subset of MITF target genes (Supplementary Fig. 21c-d, Supplementary 248

Table 12). 249

MITF is an essential melanocyte survival gene, and therefore we could not directly test if 250

MITF mediates PARP1 senescence rescue by knocking down MITF in our system. While MITF 251

depletion by shRNA dramatically reduces melanocyte growth approximately two weeks 252

following knockdown (Fig. 6c,d), at an earlier time point (D9 after infection), we noticed that 253

MITF depletion itself introduced a weak but distinct senescence-associated H3K9Me3 254

phenotype (Fig. 6e) as previously reported 42. Interestingly, in the context of MITF knockdown, 255

PARP1 senescence rescue is not observed in terms of partial reversal of H3K9Me3 focus 256

formation (Fig. 6e). These results suggest that PARP1 function in senescence rescue is likely 257

upstream of MITF expression. 258

Given that PARP1 activates MITF at the mRNA level, we performed ChIP to determine 259

whether PARP1 directly localizes to the melanocyte-specific MITF-M 43,44 promoter in human 260

primary melanocytes. ChIP experiments were performed both via overexpression of PARP1 in 261

melanocytes, as well as in melanocytes expressing endogenous levels. Among seven primer 262

sets spanning the MITF-M promoter region (–1263 to +172bp of the transcription start site; 263

TSS), the two that are more proximal to the TSS displayed the most significant enrichment, 264

indicating PARP1 binding (~8-19 fold above IgG background at endogenous level, Primer4 and 265

Primer5, from -587 to -136 bp; Fig. 7). Localization of PARP1 to the MITF-M promoter was 266

observed both with PARP1 over-expression as well as endogenous levels, while no detectable 267

enrichment was observed when PARP1 was knocked down using an shRNA (data not shown). 268

Notably, the region where PARP1 localizes overlaps with consensus binding sites of previously 269

known transcriptional regulators of the MITF-M promoter: SOX10, TCF/LEF, and CRE (Fig. 7); 270

and a putative PARP1 binding motif45 was also predicted in this core region (see Methods 271

section). 272

To assess whether PARP1 directly regulates MITF transcription, we performed 273

luciferase reporter assays using constructs containing the MITF-M promoter regions46. Altered 274

PARP1 levels via over-expression or shRNA-mediated knockdown failed to modulate reporter 275

activity (Supplementary Fig. 22), suggesting a potential epigenetic role for PARP1 in regulation 276

of the MITF-M locus that is not recapitulated via ectopic expression of a reporter gene. Bisulfite 277

sequencing of two assayable CpGs near the MITF-M TSS47 indicated that introduction of 278

BRAFV600E does not alter their methylation status (Supplementary Fig. 23), speaking against 279

DNA methylation-mediated regulation in this context. Examination of a promoter histone mark, 280

H3K4Me3, by ChIP revealed that H3K4Me3 is enriched in the MITF-M proximal promoter when 281

PARP1 is expressed either endogenously or ectopically (Supplementary Fig. 24a), but is 282

diminished upon knocking down PARP1 expression (Supplementary Fig. 24b). Furthermore, 283

expression of BRAFV600E resulted in diminished H3K4Me3 signal in the MITF-M promoter, but 284

was restored to the usual high levels upon co-expression with PARP1 (Supplementary Fig. 285

24c). ChIP analyses using antibodies recognizing the C-terminal domain (CTD) of RNA 286

polymerase II (RNA Pol II) or CTD with phosphorylated Serine 5 demonstrated that RNA Pol II 287

and its initiation-signature, Serine 5 phosphorylation, are both enriched near M-MITF TSS at 288

endogenous PARP1 levels, but diminished upon PARP1 knock-down (Supplementary Fig. 25). 289

These results are consistent with a model in which PARP1 influences MITF-M promoter activity 290

and influences MITF levels via transcriptional regulation. 291

292

DISCUSSION 293

In this study, eQTL and ASE analyses suggest PARP1 as the susceptibility gene 294

underlying the melanoma risk locus on chromosome band 1q42.1. When we evaluated the set 295

of genes in +/- 1Mb of the lead melanoma risk SNP (rs3219090) to account for potential long-296

range regulation, we observed a highly-reproducible eQTL with PARP1, but not with other 297

nearby genes. The correlation between the risk allele and higher levels of PARP1 expression 298

was highly reproducible across multiple melanoma-relevant tissues, including early-passage 299

melanoma cell lines, melanoma tumors, and human skin biopsies in both eQTL and ASE 300

analyses. While eQTL and ASE analyses cannot completely rule out a potential role for other 301

genes within the larger genomic region surrounding the GWAS peak, these data strongly 302

implicate PARP1 as functionally mediating melanoma risk at this locus. 303

While this region is relatively small in size, 65 variants are nonetheless strongly 304

correlated (r2>0.6) with the lead GWAS SNP. To efficiently prioritize functional candidates we 305

took advantage of potential gene regulatory regions annotated in human melanocyte and 306

melanoma samples by the ENCODE and Roadmap Projects. We chose to focus on variants 307

located in most consistently annotated regulatory elements across different individuals and 308

cellular conditions because of the strikingly consistent eQTL and ASE data observed in both 309

melanocytes and melanomas. Subsequent characterization of these candidate variants 310

highlighted a single variant, rs144361550, as a strong functional candidate. Of the variants 311

tested, only rs144361550 demonstrated both allele-specific transcriptional activity and protein 312

binding pattern in a manner consistent with the observed pattern of genotype/expression 313

correlation. While these data provide support for rs144361550 as a functional melanoma risk 314

variant influencing levels of PARP1 expression, they nonetheless cannot rule out other variants 315

in this region as also contributing to the observed correlation between PARP1 levels and 316

genotype. 317

Our unbiased approach using quantitative mass-spectrometry identified RECQL as a 318

protein binding allele-preferentially to rs144361550. Importantly, RECQL binding to 319

rs144361550 does not appear to be driven by sequence specificity but rather by DNA 320

secondary structure. While genomic sequence encompassing rs144361550 suggested G4-321

forming potential (Supplementary Table 11), which might explain a regulatory role48, our in 322

vitro assays failed to provide definitive evidence for G4 structure either by insertion or deletion 323

allele. However, formation of another differential structural motif, such as a transient hairpin 324

structure (formed by single-stranded sequences, Supplementary Table 13) or a locally 325

perturbed double-helix structure at the hexanucleotide repeat domain49, inducing DNA bending 326

and serving as a recognition motif for allele-specific protein binding50, cannot be excluded. 327

Although PARP1 is most well-known for its role in DNA repair, to date GWAS have 328

identified this region only for melanoma susceptibility (GWAS catalog; see URL section), 329

suggesting a potential melanoma-specific role of PARP1. Indeed, ectopic expression of PARP1 330

led to increased melanocyte proliferation both alone, as well as in the presence of oncogenic 331

mutated version of BRAF (BRAFV600E). BRAFV600E is found in most melanocytic nevi51, and as 332

reported earlier31 its over-expression in primary melanocytes led to markedly decreased cell 333

proliferation as well as H3K9Me3 focus formation, a hallmark of cellular senescence52. Our data 334

show that these senescence-like phenotypes are partially reversed by elevated PARP1 levels. 335

Further, ectopic expression of PARP1 in TERT-immortalized (p’mel) melanocytes expressing 336

BRAFV600E 35 led to a tumorigenic phenotype exemplified by anchorage-independent growth. 337

These data suggest a role for increased PARP1 in the early stages of melanomagenesis 338

promoting increased melanocyte growth and/or escape from BRAFV600E-induced senescence. In 339

this context, PARP1 seems to be functionally linked to melanocyte lineage survival oncogene, 340

MITF, which itself is an established melanoma susceptibility gene53,54. While we did not 341

exhaustively explore all the major functions of PARP1 including DNA repair and inflammation in 342

the context of melanomagenesis, our findings highlight a novel role for PARP1 in a melanocyte-343

specific context, providing a potential link between a susceptibility gene and a lineage-specific 344

oncogene. 345

Most well-established functions of PARP1 are dependent on the poly-ADP ribosylation 346

activity including its role in DNA damage response and repair. While some transcriptional co-347

regulator functions of PARP1 do not require its enzymatic activity24,25, the role of PARP1 in 348

chromatin structure modification mainly relies on NAD+-dependent association with 349

nucleosomes55 or PARylation of histones and chromatin modulators such as histone 350

demethylase KDM5B56,57. Notably, the effects of PARP1 expression on melanocyte 351

proliferation, BRAFV600E-induced senescence, and restoring MITF-M expression do not appear 352

to depend on the catalytic activity of PARP1. Interestingly, PARP1-mediated MITF-M up-353

regulation seems to involve chromatin modulation, as the promoter histone mark H3K4Me3 is 354

markedly enriched near the TSS of MITF-M upon PARP1 expression relative to its depleted 355

state in BRAFV600E-expressing melanocytes. While this is an important observation of potential 356

PARylation-independent chromatin modulation function, further investigation is required to 357

establish a causal relationship and molecular link. 358

359

URLs. 360

https://mathgen.stats.ox.ac.uk/impute/impute_v2.html 361

https://mathgen.stats.ox.ac.uk/genetics_software/snptest/snptest.html 362

http://firebrowse.org/ 363

https://tcga-data.nci.nih.gov/tcga/ 364

http://www.bios.unc.edu/research/genomic_software/Matrix_eQTL/ 365

http://www.gtexportal.org/home/testyourown 366

www.broadinstitute.org/cancer/cga/gistic 367

http://www.cbioportal.org/ 368

https://www.encodeproject.org/ 369

http://www.roadmapepigenomics.org/ 370

http://genome.ucsc.edu/ 371

http://liulab.dfci.harvard.edu/MACS/ 372

http://splice.uwo.ca/ 373

http://bioinformatics.ramapo.edu/QGRS/analyze.php 374

http://unafold.rna.albany.edu/?q=mfold 375

https://www.ebi.ac.uk/gwas/ 376

377

ACKNOWLEDGEMENTS 378

The results shown here are in part based upon data generated by the TCGA Research Network: 379

http://cancergenome.nih.gov/. Data were also obtained from the GTEx Portal on 12/02/2015 or 380

dbGaP accession number phs000424.v6.p1 on 12/17/2015. We would like to thank Dr. Hans 381

Widlund for providing p’mel cells and technical assistance, Dr. David Fisher for providing 382

luciferase construct pMITF-382 and experimental advice, Dr. William Pavan for providing 383

luciferase construct pMITF-2256, Dr. Meenhard Herlyn for providing pLU-TCMV-FMCS-pPURO 384

vector, Dr. Göran Jönsson for assistance with assessment of MITF methylation, the Arizona 385

State University sequencing and genotyping center, Christopher Hautman, Casey Dagnall, 386

Kristine Jones, and Dr. Charles Chung at The National Cancer Institute Cancer Genomics 387

Research Laboratory (CGR), Dr. Daniel Peeper for providing the lentiviral expression vector for 388

BRAFV600E constructs, Marie Webster and Ashani Weeraratna at the Wistar Institute Melanoma 389

Research Center, Dr. Stephen Chanock, Dr. Michael Dean, Dr. Laufey Amundadottir, Dr. Jason 390

Hoskins, Dr. Leandro Colli, Andrew Vu, and Christine Lee from the National Cancer Institute 391

Laboratory of Translational Genomics, and David Youngkin. This work has been supported by 392

the Intramural Research Program (IRP) of the Division of Cancer Epidemiology and Genetics, 393

National Cancer Institute, National Institutes of Health, U.S. The content of this publication does 394

not necessarily reflect the views or policies of the Department of Health and Human Services, 395

nor does mention of trade names, commercial products, or organizations imply endorsement by 396

the U.S. Government. S.M. was supported by an Australian Research Council Future 397

Fellowship, and NKH was supported by a fellowship from the National Health and Medical 398

Research Council of Australia. M.V. was supported by the Netherlands Organization for 399

Scientific Research (NWO Gravitation Program Cancer Genomics Netherlands). M.M. was 400

supported by a grant from the Marie Curie Initial Training Network (ITN) DevCom (FP7, grant 401

number 607142). M.M.I., J.A.N-B., and D.T.B. were supported by CRUK programme. We would 402

like to thank the contribution of GenoMEL consortium. 403

404

AUTHOR CONTRIBUTIONS 405

J.C., M.X., and K.M.B. designed the study. J.C., M.M.M., M.A.K., and W.J.K. conducted 406

experiments for molecular characterization of PARP1 risk variants. M.X. performed phenotypic 407

analyses of PARP1 in primary and immortalized melanocytes. Proteomics analysis was 408

conducted by M.M.M. and M.V. CD and TDS analysis was performed by A.G. and M.T. Data 409

was analyzed by T.Z., M.H.L., H.P., and M.M.I. Fine mapping of GWAS data was performed by 410

M.M.I., D.T.B., J.A.N-B., S.M., and M.H.L. Melanoma cell line eQTL and ASE experiments were 411

performed by K.M.B., N.K.H., J.M.T., M.G., and J.C. The manuscript was written by J.C., M.X., 412

and K.M.B. 413

414

COMPETING FINANCIAL INTERESTS 415

The authors declare no competing financial interests. 416

417

REFERENCES 418

1. Amos, C.I. et al. Genome-wide association study identifies novel loci predisposing to cutaneous 419

melanoma. Human molecular genetics 20, 5012-23 (2011). 420

2. Barrett, J.H. et al. Genome-wide association study identifies three new melanoma susceptibility 421

loci. Nature genetics 43, 1108-13 (2011). 422

3. Barrett, J.H. et al. Fine mapping of genetic susceptibility loci for melanoma reveals a mixture of 423

single variant and multiple variant regions. Int J Cancer 136, 1351-60 (2015). 424

4. Bishop, D.T. et al. Genome-wide association study identifies three loci associated with 425

melanoma risk. Nat Genet 41, 920-5 (2009). 426

5. Brown, K.M. et al. Common sequence variants on 20q11.22 confer melanoma susceptibility. 427

Nature genetics 40, 838-40 (2008). 428

6. Iles, M.M. et al. A variant in FTO shows association with melanoma risk not due to BMI. Nat 429

Genet 45, 428-32, 432e1 (2013). 430

7. Law, M.H. et al. Genome-wide meta-analysis identifies five new susceptibility loci for cutaneous 431

malignant melanoma. Nat Genet 47, 987-95 (2015). 432

8. Macgregor, S. et al. Genome-wide association study identifies a new melanoma susceptibility 433

locus at 1q21.3. Nature genetics 43, 1114-8 (2011). 434

9. Rafnar, T. et al. Sequence variants at the TERT-CLPTM1L locus associate with many cancer types. 435

Nat Genet 41, 221-7 (2009). 436

10. Pena-Chilet, M. et al. Genetic variants in PARP1 (rs3219090) and IRF4 (rs12203592) genes 437

associated with melanoma susceptibility in a Spanish population. BMC Cancer 13, 160 (2013). 438

11. Davies, J.R. et al. Inherited variation in the PARP1 gene and survival from melanoma. Int J Cancer 439

135, 1625-33 (2014). 440

12. Law, M.H. et al. PARP1 polymorphisms play opposing roles in melanoma occurrence and 441

survival. Int J Cancer 136, 2488-9 (2015). 442

13. Krishnakumar, R. & Kraus, W.L. The PARP side of the nucleus: molecular actions, physiological 443

outcomes, and clinical targets. Mol Cell 39, 8-24 (2010). 444

14. Woodhouse, B.C. & Dianov, G.L. Poly ADP-ribose polymerase-1: an international molecule of 445

mystery. DNA Repair (Amst) 7, 1077-86 (2008). 446

15. Huber, A., Bai, P., de Murcia, J.M. & de Murcia, G. PARP-1, PARP-2 and ATM in the DNA damage 447

response: functional synergy in mouse development. DNA Repair (Amst) 3, 1103-8 (2004). 448

16. Bouchard, V.J., Rouleau, M. & Poirier, G.G. PARP-1, a determinant of cell survival in response to 449

DNA damage. Exp Hematol 31, 446-54 (2003). 450

17. Maynard, S., Schurman, S.H., Harboe, C., de Souza-Pinto, N.C. & Bohr, V.A. Base excision repair 451

of oxidative DNA damage and association with cancer and aging. Carcinogenesis 30, 2-10 (2009). 452

18. Swindall, A.F., Stanley, J.A. & Yang, E.S. PARP-1: Friend or Foe of DNA Damage and Repair in 453

Tumorigenesis? Cancers (Basel) 5, 943-58 (2013). 454

19. Urabe, K. et al. The inherent cytotoxicity of melanin precursors: a revision. Biochim Biophys Acta 455

1221, 272-8 (1994). 456

20. Kuilman, T., Michaloglou, C., Mooi, W.J. & Peeper, D.S. The essence of senescence. Genes Dev 457

24, 2463-79 (2010). 458

21. Leikam, C., Hufnagel, A., Schartl, M. & Meierjohann, S. Oncogene activation in melanocytes links 459

reactive oxygen to multinucleated phenotype and senescence. Oncogene 27, 7070-82 (2008). 460

22. Kraus, W.L. Transcriptional control by PARP-1: chromatin modulation, enhancer-binding, 461

coregulation, and insulation. Curr Opin Cell Biol 20, 294-302 (2008). 462

23. Schiewer, M.J. & Knudsen, K.E. Transcriptional roles of PARP1 in cancer. Mol Cancer Res 12, 463

1069-80 (2014). 464

24. Hassa, P.O. & Hottiger, M.O. The functional role of poly(ADP-ribose)polymerase 1 as novel 465

coactivator of NF-kappaB in inflammatory disorders. Cell Mol Life Sci 59, 1534-53 (2002). 466

25. Cervellera, M.N. & Sala, A. Poly(ADP-ribose) polymerase is a B-MYB coactivator. J Biol Chem 275, 467

10692-6 (2000). 468

26. Consortium, E.P. An integrated encyclopedia of DNA elements in the human genome. Nature 469

489, 57-74 (2012). 470

27. Roadmap Epigenomics, C. et al. Integrative analysis of 111 reference human epigenomes. 471

Nature 518, 317-30 (2015). 472

28. Boersema, P.J., Raijmakers, R., Lemeer, S., Mohammed, S. & Heck, A.J. Multiplex peptide stable 473

isotope dimethyl labeling for quantitative proteomics. Nat Protoc 4, 484-94 (2009). 474

29. Hubner, N.C., Nguyen, L.N., Hornig, N.C. & Stunnenberg, H.G. A quantitative proteomics tool to 475

identify DNA-protein interactions in primary cells or blood. J Proteome Res 14, 1315-29 (2015). 476

30. Butter, F. et al. Proteome-wide analysis of disease-associated SNPs that show allele-specific 477

transcription factor binding. PLoS Genet 8, e1002982 (2012). 478

31. Michaloglou, C. et al. BRAFE600-associated senescence-like cell cycle arrest of human naevi. 479

Nature 436, 720-4 (2005). 480

32. Kaplon, J. et al. A key role for mitochondrial gatekeeper pyruvate dehydrogenase in oncogene-481

induced senescence. Nature 498, 109-12 (2013). 482

33. Vredeveld, L.C. et al. Abrogation of BRAFV600E-induced senescence by PI3K pathway activation 483

contributes to melanomagenesis. Genes Dev 26, 1055-69 (2012). 484

34. Zhuang, D. et al. C-MYC overexpression is required for continuous suppression of oncogene-485

induced senescence in melanoma cells. Oncogene 27, 6623-34 (2008). 486

35. Garraway, L.A. et al. Integrative genomic analyses identify MITF as a lineage survival oncogene 487

amplified in malignant melanoma. Nature 436, 117-22 (2005). 488

36. Eltze, T. et al. Imidazoquinolinone, imidazopyridine, and isoquinolindione derivatives as novel 489

and potent inhibitors of the poly(ADP-ribose) polymerase (PARP): a comparison with standard 490

PARP inhibitors. Mol Pharmacol 74, 1587-98 (2008). 491

37. Wacker, D.A. et al. The DNA binding and catalytic domains of poly(ADP-ribose) polymerase 1 492

cooperate in the regulation of chromatin structure and transcription. Mol Cell Biol 27, 7475-85 493

(2007). 494

38. Pavri, R. et al. PARP-1 determines specificity in a retinoid signaling pathway via direct 495

modulation of mediator. Mol Cell 18, 83-96 (2005). 496

39. Rolli, V., O'Farrell, M., Menissier-de Murcia, J. & de Murcia, G. Random mutagenesis of the 497

poly(ADP-ribose) polymerase catalytic domain reveals amino acids involved in polymer 498

branching. Biochemistry 36, 12147-54 (1997). 499

40. Haq, R. et al. Oncogenic BRAF regulates oxidative metabolism via PGC1alpha and MITF. Cancer 500

Cell 23, 302-15 (2013). 501

41. Muller, J. et al. Low MITF/AXL ratio predicts early resistance to multiple targeted drugs in 502

melanoma. Nat Commun 5, 5712 (2014). 503

42. Giuliano, S. et al. Microphthalmia-associated transcription factor controls the DNA damage 504

response and a lineage-specific senescence program in melanomas. Cancer Res 70, 3813-22 505

(2010). 506

43. Huber, W.E. et al. A tissue-restricted cAMP transcriptional response: SOX10 modulates alpha-507

melanocyte-stimulating hormone-triggered expression of microphthalmia-associated 508

transcription factor in melanocytes. J Biol Chem 278, 45224-30 (2003). 509

44. Takeda, K. et al. Induction of melanocyte-specific microphthalmia-associated transcription factor 510

by Wnt-3a. J Biol Chem 275, 14013-6 (2000). 511

45. Ko, H.L. & Ren, E.C. Functional Aspects of PARP1 in DNA Repair and Transcription. Biomolecules 512

2, 524-48 (2012). 513

46. Watanabe, A., Takeda, K., Ploplis, B. & Tachibana, M. Epistatic relationship between 514

Waardenburg syndrome genes MITF and PAX3. Nat Genet 18, 283-6 (1998). 515

47. Lauss, M. et al. Genome-Wide DNA Methylation Analysis in Melanoma Reveals the Importance 516

of CpG Methylation in MITF Regulation. J Invest Dermatol 135, 1820-8 (2015). 517

48. Bochman, M.L., Paeschke, K. & Zakian, V.A. DNA secondary structures: stability and function of 518

G-quadruplex structures. Nat Rev Genet 13, 770-80 (2012). 519

49. Stefl, R. et al. A-like guanine-guanine stacking in the aqueous DNA duplex of d(GGGGCCCC). J 520

Mol Biol 307, 513-24 (2001). 521

50. Lu, X.J., Shakked, Z. & Olson, W.K. A-form conformational motifs in ligand-bound DNA 522

structures. J Mol Biol 300, 819-40 (2000). 523

51. Pollock, P.M. et al. High frequency of BRAF mutations in nevi. Nat Genet 33, 19-20 (2003). 524

52. Narita, M. et al. Rb-mediated heterochromatin formation and silencing of E2F target genes 525

during cellular senescence. Cell 113, 703-16 (2003). 526

53. Yokoyama, S. et al. A novel recurrent mutation in MITF predisposes to familial and sporadic 527

melanoma. Nature 480, 99-103 (2011). 528

54. Bertolotto, C. et al. A SUMOylation-defective MITF germline mutation predisposes to melanoma 529

and renal carcinoma. Nature 480, 94-8 (2011). 530

55. Krishnakumar, R. et al. Reciprocal binding of PARP-1 and histone H1 at promoters specifies 531

transcriptional outcomes. Science 319, 819-21 (2008). 532

56. Kim, M.Y., Zhang, T. & Kraus, W.L. Poly(ADP-ribosyl)ation by PARP-1: 'PAR-laying' NAD+ into a 533

nuclear signal. Genes Dev 19, 1951-67 (2005). 534

57. Krishnakumar, R. & Kraus, W.L. PARP-1 regulates chromatin structure and transcription through 535

a KDM5B-dependent pathway. Mol Cell 39, 736-49 (2010). 536

537

Figure 1 The melanoma risk-associated G allele of rs3219090 is correlated with increased PARP1 538

expression. (a) eQTL analysis was performed for rs3219090 using expression microarray and SNP array 539

genotypes derived from a panel of 59 early-passage melanoma cell lines. A significant eQTL was 540

observed for PARP1, and the result is plotted for rs3219090 genotype (P = 1.4 x 10-3; linear regression). 541

G is the risk allele and A the protective allele of rs3219090. A.U.; arbitrary unit. (b) The allelic ratios of 542

PARP1 transcripts were measured in 14 copy-neutral melanoma cell lines that were heterozygous for 543

both rs3219090 and a synonymous mRNA-coding surrogate SNP (rs1805414, r2=0.98 with rs3219090) 544

using Taqman genotyping assays. Allelic ratios were inferred from a known amount of allelic standards 545

and plotted as a ratio of PARP1 expression from the risk over protective allele (P = 1.2 x 10-4, two-tailed 546

Wilcoxon signed rank test, average value of PCR triplicates were considered as a single data point). (c) 547

Allelic ratios of PARP1 transcripts were measured using RNA sequencing data from 48 copy-neutral 548

TCGA skin melanoma samples that were heterozygous for both rs3219090 and rs1805414. The mapped 549

numbers of RNAseq reads encompassing each allele of rs1805414 were used for calculating allelic ratios 550

(P = 0.011, two-tailed Wilcoxon signed rank test). Solid line marks 1:1 ratio, and dashed line represents 551

median ratio. 552

553

Figure 2 Functional annotation of a 3kb region encompassing rs144361550 in primary melanocytes. 554

Histone modifications (H3K4Me1, H3K4Me3, and H3K27Ac) and DNaseI hypersensitivity sites (DHS) in 555

primary melanocytes are shown for a 3kb region encompassing rs144361550. The red dashed vertical 556

line indicates the position of rs144361550, overlapping histone marks, DHS, and transcription factor 557

binding sites. Genomic positions are based on hg19. Transcription factor binding sites are from UCSC 558

genome browser track “Transcription Factor ChIP-seq (161 factors) from ENCODE with Factorbook 559

Motifs” representing multiple ENCODE cell types. “Chromatin Primary Core Marks Segmentation by HMM 560

from Roadmap Project” track is also shown for three melanocyte samples. TssA: Active_TSS, TssF: 561

Flanking_Active_TSS, Enh: Enhancers. For DHS, traces from two experimental replicates of Melanocyte 562

1 and 2 are displayed. The scale of each track is uniformly set throughout the region of the PARP1 gene 563

to cover the highest peaks, with 0 as the baseline (see online methods for details of each track). 564

565

Figure 3 The melanoma-associated indel, rs144361550, drives allelic transcriptional activity and protein 566

binding. (a,c) Luciferase assays for rs144361550 were conducted using the melanoma cell line 567

UACC2331 (a) or primary melanocytes (c). 905bp or 22/28bp encompassing rs144361550, respectively, 568

were cloned 5’ of minimal promoter in pGL4.23 vector and transfected into the cells. Luciferase activity 569

was measured 24hrs after transfection and was normalized against Renilla luciferase activity. Relative 570

luciferase levels were plotted as percent of the minimal promoter control. P-values are shown or * P < 571

0.05 against Ctrl (two-tailed, paired t-test). Two (a) and three (c) independent cell cultures and 572

transfections of n = 3 were combined to present the median with range, 75 & 25 percentiles, and each 573

data point. Ctrl: minimal promoter control, Del: deletion/risk allele construct, Ins: insertion/protective allele 574

construct. (b,d) EMSA was performed using biotin-labeled double stranded oligos for the deletion/risk 575

(Del, D) or insertion/protective (Ins, I) alleles of rs144361550 and nuclear extracts from melanoma cell 576

lines UACC2331 and UACC457 (b) or primary melanocytes (d). Sequences shown in panel b for the 22 577

bp deletion and 28 bp insertion probes were used for both b and d (bold and Italic bases highlight 578

potential G4 structure forming nucleotides). 50X, 200X or 500X molar excess of unlabeled competitors 579

were added in specified lanes. 580

Figure 4 RECQL binds to the insertion allele and mediates allelic expression. (a) Insertion allele-specific 581

binding proteins were identified by mass-spectrometry using melanoma cell nuclear extract and 582

biotinylated double-stranded oligos. The ratio of proteins bound to heavy/light-dimethyl labeled probes is 583

plotted on x and y-axis for labeling swapping. Red circles: enriched above the background in both 584

directions. Circle sizes represent relative abundance. (b) RECQL EMSA/supershift. D: deletion, I: 585

insertion (c) ChIP was performed using anti-RECQL antibody or IgG and melanoma cell chromatin 586

followed by qPCR. DNA quantity was normalized to input DNA for each IP (n=3). Neg: gene desert, Pos: 587

a known RECQL binding locus, Indel: rs144361550 region. A representative set from four independent 588

experiments is shown. (d) RECQL under tetracycline-inducible promoter was expressed in three 589

melanoma cell lines. PARP1 levels (top) and RECQL RNA (middle) and protein (bottom) levels were 590

measured at 48hrs of doxycycline induction (blot images were cropped). Transcript levels are shown as 591

fold over Empty vector after normalizing to B2M control (n = 6, 5, and 6 for each cell line). (e) Luciferase 592

assays were performed using 905bp deletion (Del) or insertion allele (Ins) constructs with RECQL or 593

Empty vector co-transfection in HEK293FT cells. Renilla-normalized relative luciferase activities were 594

plotted as percent of the minimal promoter control (Ctrl) (n = 6). (c-e) Each graph shows median with 595

range, 75 & 25 percentiles, and each data point. Two-tailed, t-test assuming unequal variance for all P-596

values shown. * P < 0.05 against Ctrl (e) or Empty (d). 597

598

Figure 5 Cell growth and H3K9Me3 focus formation in primary human melanocytes expressing PARP1 599

and BRAFV600E. (a) Crystal violet (CV) staining of melanocytes expressing PARP1, BRAFV600E, or both. 600

Cells were first infected with pIN20-PARP1 followed by G418 treatment and doxycycline induction. Cells 601

were then infected with HIV-CSCG-BRAFV600E vector at day 6 after PARP1 infection followed by 602

blasticidine selection. Equal cell numbers were seeded at day 11 after pIN20-PARP1 infection and 603

stained with CV 9 days after seeding. (b) BrdU staining of primary human melanocytes expressing either 604

PARP1, BRAFV600E, or both. BrdU staining was done at day 13 after initial infection of pIn20-PARP1 (n = 605

4, median with range and 25 & 75 percentiles, t-test assuming unequal variance, representative set from 606

three experiments). (c) H3K9Me3 staining of primary human melanocytes at day 13 after initial infection. 607

Scale bars are shown for 10µm. (d) Colony formation in soft-agar of p’mel/BRAFV600E cells expressing 608

either PARP1 or MITF-M. Immortalized melanocytes (p’mel cells) expressing active BRAFV600E were 609

infected with pLX304 empty, pLX304-PARP1 or pLX304-MITF and seeded in soft-agar plates after 610

blasticidine selection. Images were taken at day 22 and colonies were counted from two sets of 611

triplicates at day 26 after plating (n=6, from two experiments, median with range and 25 & 75 percentiles). 612

t-test assuming unequal variance (e) Representative images of colonies are shown. The same 613

magnification (40X) was applied for all the images. 614

615

Figure 6 MITF expression is restored in primary human melanocytes co-expressing PARP1 and 616

BRAFV600E in a PARylation-independent manner, concurrent with partial reversal of senescence 617

phenotypes. (a) Western blot detection of MITF (long and short exposures), BRAF, PARP1, and PAR. 618

Primary human melanocytes were co-infected with PARP1, BRAFV600E followed by treatment with the 619

PARP inhibitor, BYK204165. Cell extract was generated at day 13 after initial infection with PARP1 620

construct. Blot images were cropped. (b) BrdU flow cytometry of primary human melanocytes co-infected 621

with PARP1, BRAFV600E, and/or MITF-shRNA followed by treatment with the PARP inhibitor, BYK204165. 622

(c) Crystal Violet staining of melanocytes co-infected with PARP1, BRAFV600E, and/or MITF-shRNA. 623

Primary melanocytes were first infected with pIN20-PARP1 or empty pIN20 vector concurrently with 624

pLKO-shRNA-MITF-M or pLKO-empty vector followed by G418 treatment and doxycycline induction. 625

Cells were then infected with HIV-CSCG-BRAFV600E or empty HIV-CSCG vector at day 6 after PARP1 626

infection followed by Blasticidine selection. A representative crystal violet (CV) staining (n=3) image is 627

displayed at day 17 of infection. (d) Quantification of solubilized CV staining by measuring absorbance at 628

540nm at day 17 of infection is plotted. Median with range (n=3, t-test assuming unequal variance, a 629

representative set from three experiments is shown) (e) PARP1-mediated partial reversal of BRAFV600E-630

induced H3K9Me3 focus formation in melanocytes is blocked by knockdown of MITF. H3K9Me3 staining 631

was performed at day 13 after initial infection. Scale bars are shown for 10µm. 632

633

Figure 7 PARP1 binds to MITF-M promoter. Human melanocytes were transduced with lentiviral 634

constructs of empty pIN20 (Empty = endogenous PARP1 levels) or PARP1 cDNA (PARP1 = over-635

expression of PARP1) and harvested five days after infection. Chromatin immunoprecipitation was 636

performed using antibody against PARP1 (anti-PARP1) or rabbit normal IgG (IgG) and pulled-down DNA 637

was amplified using seven sets of primers spanning -1263bp through to +172bp relative to MITF-M TSS 638

(black arrow). Gray arrows depict locations of qPCR primers relative to the genomic region of MITF-M 639

promoter shown below. Relative quantities were calculated by normalizing each sample amount to 640

matched input DNA. Average quantity with range, and individual data points for qPCR triplicates are 641

shown. Experiments were repeated 3 times yielding similar results, and the results from a representative 642

set are displayed. Known (closed symbols) or predicted (open symbols) binding sites for transcription 643

factors are presented to their relative genomic positions on MITF-M promoter. The genomic sequence of -644

325bp through to -150bp relative to the TSS are shown in the box, with known or predicted transcription 645

factor binding sites underlined. 646

ONLINE METHODS 647

Early passage melanoma cell line eQTL analysis. Early passage melanoma cell lines were 648

obtained from the University of Arizona Cancer Center (UACC), and eQTL analysis was 649

performed by combining gene expression profiling and SNP genotyping data. The use of cell 650

lines was approved by National Institutes of Health Office of Human Subject Research. Early 651

passage melanoma cell lines were grown in the medium containing RPMI1640, 10% FBS, 20 652

mM HEPES, and penicillin/streptomycin until ~70% confluent. All cell lines were tested negative 653

for mycoplasma contamination. For RNA isolation, cells were washed twice with cold PBS on 654

ice and lysed with Trizol. Trizol was heated to 65ºC for 5 min to maximize melanin removal. 655

Following heating, 1 mL chloroform was added per 5 mL of Trizol, vortexed, cooled on ice for 5 656

min, and centrifuged. The aqueous phase was removed, and equal volume of 70% EtOH was 657

added dropwise while vortexing at low speed. Ethanol /supernatant mixtures were added to 658

Qiagen RNeasy midi columns, with the flow-through reapplied once. Samples were then 659

processed per manufacturer’s protocol. RNA quantity and integrity were assessed using 660

Bioanalyzer, which yielded RIN>7 for all samples. Total RNA were expression profiled on 661

Affymetrix U133Plus2 expression microarrays, with labeling, hybridization, washing, and 662

scanning performed according to manufacturer’s protocol. Background correction and quantile 663

normalization of gene expression data were performed using Robust Multi-array Average (RMA) 664

algorithm with the default settings (Affymetrix). For genomic DNA isolation, Qiagen DNeasy 665

Blood and Tissue kit was used. DNA quantity was measured using NanoDrop and PicoGreen 666

fluorescent assay. All samples were profiled using Applied Biosystems Identifiler STR panel 667

prior to genotyping on Illumina OmniExpress arrays. After quality assessment of genotypes 668

samples with >0.1 missing rate were excluded from the analysis. Loci with > 0.1 missing rate, 669

MAF < 0.01, or Hardy-Weinberg Equilbrium P- value < 5E-5 were also excluded. The genomic 670

region encompassing 6Mb around the GWAS lead SNP rs3219090 (which was directly 671

genotyped on the array) was imputed using IMPUTE2.2.258 and 1KG phase1 v3 April 2012 672

(build 37) as a reference data. After assigning imputed genotypes for 2 samples that were 673

missing direct genotype for rs3219090 (recoded as 0.333 probability of each genotype), 59 total 674

samples were qualified for eQTL analysis with gene expression and genotype data available. 675

Affymetrix U133Plus2 annotates 17 genes and 11 other transcripts in the 2 Mb region centering 676

at rs3219090. Among these, probes for 12 genes and 2 other transcripts passed QC, including 677

PARP1. eQTL analysis was then performed for these samples and gene/transcripts using 678

SNPTEST v2.5 (see URL section) considering an additive model for genotypes. 679

Allele discrimination qPCR. cDNA from early passage melanoma cell lines heterozygous for 680

both rs3219090 and coding surrogate SNP rs1805414 as well as of normal genomic copy were 681

assayed using custom-designed Taqman genotyping probe sets that do not recognize genomic 682

DNA. To act as a standard, the same amplicon was PCR-amplified for each allele from cDNA 683

and subsequently cloned into the pCR2.1 TOPO vector (Invitrogen) and sequence verified. A 684

standard curve was then generated using known amounts of cloned amplicons by plotting 11 685

different points of allelic ratio against VIC/FAM signal ratio. Allele discrimination qPCR was 686

performed in triplicate, and allelic ratio was calculated from the average ratio of VIC/FAM signal 687

using the standard curve. Departure from expected allelic ratio (major/minor allele) of 1 was 688

assessed using two-tailed Wilcoxon signed rank test. 689

Nomination of candidate functional variants. All LD r2 values used for candidate variant 690

nomination were from 1KG phase 3 data. r2 values based on both the EUR and CEU 691

populations were considered to extract the maximum r2 of each variant with the lead SNPs, 692

rs3219090 and rs1858550. Meta-analysis P-values were obtained from the previously published 693

work of Law and colleagues7; all samples used in the meta-analysis were collected with 694

informed consent and ethics committee approvals as previously described. DHS peaks for the 695

primary human melanocyte culture “melano” (ENCODE/Duke) were obtained from ENCODE 696

database26 through UCSC Genome browser (see URL section). DNase-seq data for penis 697

foreskin melanocyte primary cell cultures “skin 01” and “skin 02” (shown as Melanocyte_1 and 698

Melanocyte_2 in Fig 2 and Supplementary Fig 2) were obtained from Roadmap database 699

(03/09/2015) and DHS peaks were called using MACS59(see URL section) using the default 700

settings and FDR 1% cutoff. DHS peak intervals were overlaid with the genomic position of 701

each candidate variant to determine whether each candidate localizes within a DHS peak. 702

Experimental duplicates for skin 01 (DS18590, DS18601) and skin 02 (DS19662, DS18668), 703

and analytical duplicates for melano were used for our analysis. To call a variant to be within 704

DHS in one sample, DHS overlapping the variant in either of the duplicates were counted. DHS 705

peaks from DNase-seq data for two melanoma cell lines Mel2183 (ENCODE/Duke) and RPMI-706

7951 (ENCODE/UW) were obtained from the ENCODE database. Peaks from FAIREseq data 707

for 11 melanoma culture samples were obtained from GEO (accession number: GSE60666). 708

Histone mark annotation was performed in the same way. Primary melanocyte histone marks 709

were taken from subsets of three individuals through Roadmap database (skin 01, skin 02, and 710

skin 03; shown as Melanocyte_1, Melanocyte_2, and Melanocyte_3 in Fig 2 and 711

Supplementary Fig 2). Source of each track visualized on UCSC genome browser is as 712

follows: Melanocyte track names used for Fig 2 (Track 1: UCSF-UBC-USC Penis Foreskin 713

Melanocyte primary Cells Histone H3K27ac Donor skin03 Library A15584 EA Release 9, Track 714

2: UCSF-UBC-USC Penis Foreskin Melanocyte primary Cells Histone H3K4me1 Donor skin03 715

Library A15579 EA Release 8, Track 3: UCSF-UBC-USC Penis Foreskin Melanocyte primary 716

Cells Histone H3K4me3 Donor skin03 Library A15580 EA Release 8, Track 4: Melano DNaseI 717

HS Density Signal from ENCODE/Duke, Track 5: Penis Foreskin Melanocyte Primary Cells 718

Donor skin 01 DNase Uniformly Signal from Roadmap, Track 6: UW Penis Foreskin Melanocyte 719

Primary Cells DNase Hypersensitivity Donor skin02 Library DNase DS19662 EA Release 9.), 720

Melanoma track names used for Supplementary Fig 2 (Track 1 through 11: H3K27ac ChIP-seq 721

signal from F-seq, Track 12 through 22: FAIRE-seq signal from F-seq, Track 23: Mel2183 722

DNaseI HS Density Signal from ENCODE/Duke, Track 24: RPMI-7951 DNaseI HS Raw Signal 723

Rep 1 from ENCODE/UW.). 724

Luciferase reporter assays. Luciferase constructs were generated to include the DHS region 725

encompassing each SNP. Sequences encompassing each variant were amplified from 726

genomic DNA of HapMap CEU panel samples using the primers listed in Supplementary Table 727

14, cloned into pCR2.1TOPO vector, and subsequently cloned into pGL4.23 vector using 728

EcoRV and HindIII enzymes or directly into pGL4.23 vector using primers with HindIII and XhoI 729

sequence overhangs. Sequence-verified pGL4.23 constructs were then co-transfected with 730

pGL4.74 (Renilla luciferase) into a melanoma cell lines UACC2331, UACC457, or UACC1308 731

using Lipofectamine 2000 reagent (Thermo Fisher) or electroporation with Lonza Amaxa SE kit 732

and DS-150 protocol on 4D-Nucleofector system. Electoporation of primary human melanocytes 733

(HEMn-LP, Invitrogen) was performed using the Lonza Amaxa P2 kit and protocol CA-137 734

(Lonza). When luciferase assays were combined with RECQL over-expression, empty pCMV6-735

Entry vector or human RECQL cDNA clone (Origene, RC200427) were co-transfected with 736

luciferase constructs into HEK293FT cells using Lipofectamine 2000 reagent. Cells were 737

collected 24hr following transfection and luciferase activity was measured using Dual-Luciferase 738

reporter system (Promega) on GLOMAX multi detection system (Promega). All cell lines and 739

primary cultures used here and onward were tested negative for mycoplasma contamination. 740

EMSA and antibody supershift. Nuclear extracts were prepared from actively growing normal 741

human melanocytes (HEMn-LP, Invitrogen or melanoma cell lines (UACC) using NE-PER 742

nuclear and cytoplasmic extraction kit (Thermo Scientific). DNA oligos for each variant were 743

synthesized with 5’ biotin labeling, and HPLC-purified (Life Technologies; probe sequences are 744

listed in Supplementary Table 14). Forward and reverse strands were then annealed to make 745

double stranded DNA probes. Probes were bound to 0.5-4µg nuclear extracts pre-incubated 746

with 1µg poly d(I-C) in binding buffer containing 10mM Tris (pH 7.5), 50 mM KCl, 1 mM DTT, 10 747

mM MgCl2, with or without 5% glycerol at 4ºC for 30min. Unlabeled competitor oligos were 748

added to the reaction mixture 5min prior to the addition of probes. Completed reactions were 749

run on 5% or 4-20% native acrylamide gel and transferred blots were developed using LightShift 750

Chemiluminescent EMSA kit (Thermo Scientific) and exposed on film. Supershift antibodies 751

(RecQL, sc-25547, Santa Cruz) or rabbit normal IgG (sc-2027, Santa Cruz) were bound to 752

nuclear extract prior to poly d(I-C) incubation at 4 ºC for 1hr. EMSAs with purified recombinant 753

protein were performed using RECQL (TP300427, Origene), where purified recombinant 754

proteins were used in place of nuclear extract and poly d(I-C). Additional antibodies and 755

recombinant proteins for validations are as follows. Antibodies are from Santa Cruz unless 756

otherwise specified: anti-NCL (sc-8031), anti-SRSF3 (sc-13510), anti-CIRP (sc-161012), anti-757

BLM (sc-7790), anti-hnRNPD (sc-22368), anti-RBM3 (sc-162080), anti-TOP3A (sc-11257), anti-758

RPA1 (sc-14696), anti-DHX36 (A300-525A, Bethyl), anti-RPA3 (ab167593, Abcam). 759

Recombinant proteins are from Origene: NCL (TP319082), CIRP (TP301639), 760

RPA1(TP302066), hnRNPD (TP300660), RBM3 (TP760298). 761

Mass-spectrometry. Quantitative AP-MS/MS following SNP DNA pulldown and in-solution 762

dimethyl chemical labeling was performed based on procedures described previously29,30. 763

Nuclear extract from the melanoma cell line UACC2331 was collected as described previously 764

using the Dignam lysis protocol60. For DNA pulldowns, 500 pmol of annealed, forward strand 5’-765

biotinylated oligo probes were coupled to streptavidin sepharose beads (GE Healthcare). 766

Insertion and deletion allele probe sequences are listed in Supplementary Table 14. Beads 767

were incubated with 450 µg of nuclear extract for 90 minutes plus 10 µg of non-specific 768

competitor DNA (either 10 µg of poly-dAdT or 5 µg of poly-dAdT plus 5 µg poly-dIdC). After 769

washes, beads were resuspended in 100mM TEAB buffer, proteins were reduced with 5mM 770

TCEP, alkylated with 10 mM MMTS, and digested overnight with 0.25 µg trypsin. Digested 771

peptides were labelled using in-solution dimethyl chemical labelling as described previously28. 772

All experiments were performed in duplicate, and labels were swapped between replicate pairs 773

to prevent labeling bias. Heavy and light labelled peptides were mixed and prepared using C18-774

StageTips. Peptides were loaded on a column packed with 1.8 µm Reprosil-Pur C18-AQ beads 775

(gift from Dr. Maisch) and eluted using a 120 minute gradient from 7%-32% buffer B (80% 776

acetonitrile, 0.1% formic acid) at a flow rate of 250nL/min. Peptides were sprayed directly onto a 777

Thermo QExactive mass spectrometer. Data was collected in top10 data-dependent acquisition 778

mode. Thermo RAW files were analyzed with MaxQuant (version 1.3.0.5) by searching against 779

the Uniprot curated human proteome. Methionine oxidation and N-terminal acetylation were 780

considered as variable modifications and cysteine-dithiomethane was set as a fixed 781

modification. Protein ratios normalized by median ratio shifting as described previously61 were 782

used for outlier calling. An outlier cutoff of 1.5 IQRs (inter-quartile ranges) in two out of two 783

biological replicates was used. 784

Chromatin immunoprecipitation of RECQL. UACC2331 melanoma cells or primary human 785

melanocytes (HEMn-LP, Invitrogen) were fixed with 1% formaldehyde when 80-90% confluent, 786

following the instructions of Active Motif ChIP-ITexpress kit or ChIP-IT high sensitivity kit. 7.5 x 787

106 cells were then sheared by sonication using a Bioruptor (Diagenode) at high setting for 788

15min, with 30 sec on and 30 sec off cycles. Sheared chromatin from 1 to 4 x 106 cells were 789

used for each immunoprecipitation with antibodies against RECQL, H110 (sd-25547; Santa 790

Cruz), and A300 (A300-450A; Bethyl), or normal rabbit IgGs (sc-2027; Santa Cruz) following the 791

manufacturer’s instructions. Purified pulled-down DNA was assayed by SYBR Green qPCR for 792

enrichment of target sites using primers listed in Supplementary Table 14. A commercial 793

primer set (71001, Active Motif) recognizing a gene desert on chromosome 12 was used for a 794

negative control (Neg). 795

Overexpression of RECQL in melanoma cell lines. RECQL was cloned from a cDNA 796

construct (RC200427, purchased from Origene) into lentiviral pLU-TCMV-FMCS-pPURO vector 797

(a generous gift from Dr. Meenhard’s lab at Wistar) containing tetracycline-inducible promoter. 798

Lentiviral vectors were co-transfected into 293 cells with packaging vectors psPAX2, pMD2-G, 799

and pCAG4-RTR2. Virus was collected two days after transfection and concentrated by 800

Vivaspin. Cells were incubated with virus for 24 hr, followed by puromycin (1-2 µg/ml) selection 801

for 2-3 days. After drug selection cells were seeded and grown in the media containing varying 802

amount of doxycycline (0,0.5, and 1 µg/ml). Cells were harvested after 48 hrs of doxycycline 803

induction for RNA and protein isolation. cDNA was generated from total RNA and transcript 804

levels were measured using Taqman qPCR. RECQL and PARP1 transcript levels normalized to 805

the levels of B2M, and PCR triplicates were averaged and considered as one data point. 806

Western blotting of RECQL and GAPDH was performed using the following antibodies: 807

RecQL1, sc-25547, Santa Cruz, GAPDH, sc-51907, Santa Cruz. 808

Lentiviral vectors and lentiviral infection of melanocytes. HIV-CSCG-BRAFV600E was used 809

to express BRAFV600E in normal human melanocytes (provided by Dr. Daniel Peeper, 810

Netherlands Cancer Institute). The PARP1 cDNA clone (BC037545) was purchased from 811

Thermo Scientific and sub-cloned and expressed in three different lentiviral vectors: a. under a 812

tetracycline-inducible promoter in the pInducer 20 vector backbone (Addgene) as pIN20-813

PARP1, which expresses high level of PARP1; b. under a tetracycline-inducible promoter in the 814

pLU-TCMV-FMCS-pPURO vector (a generous gift from Dr. Meenhard Herlyn’s laboratory at the 815

Wistar Institute) as pLU-TCMV-PARP1 which expresses low level of PARP1; c. with a CMV 816

promoter in the pLX304 backbone (Addgene) as pLX-PARP1. The PARP1 catalytic mutant 817

p.Glu988Lys39 was generated using the QuickChange II XL Site-Directed Mutagenesis kit 818

(Agilent). The MITF-M cDNA clone (NM_000248) was purchased from Origene and was sub-819

cloned and expressed in pLX304 backone as pLX-MITF. pLKO_MITF shRNAs (a pool of 4 820

shRNAs with clone IDs from TRC0000019119 to TRC0000019123) were purchased from Sigma 821

in lenitiviral-based vectors. Primary human melanocytes were obtained from Invitrogen and/or 822

the Yale SPORE in Skin Cancer Specimen Resource Core and grown under standard culture 823

conditions using Medium M254 with supplements (Invitrogen). For lentivirus production, 824

lentiviral vectors were co-transfected into 293 cells with packaging vectors psPAX2, pMD2-G, 825

and pCAG4-RTR2. Virus was collected two days after transfection and concentrated by 826

Vivaspin. Cells were incubated with virus for 24 hr, followed by drug selection, before being 827

subjected to various experimental treatments and assays. Dose of BYK204165 was chosen by 828

performing a titration assay. Briefly, melanocytes were treated with 1, 5, 20, and 40 µM 829

BYK204165 for 1, 7, and 24 hrs, and PARylation was assessed by western blotting. PARylation 830

was mostly diminished with 5 µM BYK204165, and completely undetectable with 20 µM after 1 831

hour treatment. Based on these data, we chose to use 10 µM BYK204165 for our assays. 832

Cell proliferation assays. Cell proliferation was assayed by a bromodeoxyurifine (BrdU) flow 833

kit (BD Pharmingen, San Jose, CA) according to the manufacturer’s protocol. Briefly, cells were 834

labeled with 10µM BrdU for three hours before they were fixed, permeabalized and subjected to 835

DNaseI treatment. Cells were then stained with FITC-conjugated anti-BrdU antibody and 7-836

AAD, followed by flow cytometry analysis using a FACSCalibur (BD Pharmingen, San Jose, 837

CA). For most cases, BrdU assay was performed on cells two weeks after initial infection of 838

PARP1 vector. For crystal violet (CV) staining, cells were seeded at equal numbers after 839

infection and drug selection, and stained with CV between 2-3 weeks after initial PARP1 840

infection. 841

Western and immunofluorescence staining. For western blot analysis, total cell lysates were 842

generated with RIPA buffer (Thermo Scientific, Pittsburgh, PA) and subjected to water bath 843

sonication. Samples were resolved by 4-12% Bis-Tris ready gel (Invitrogen, Carlsbad, CA) 844

electrophoresis. The primary antibodies used were rabbit anti-PAR (551813, BD Pharmingen, 845

CA), rabbit anti-PARP (9542, Cell Signaling Technology), mouse anti-MITF (MS-771-P1, 846

Thermo Scientific), mouse anti-く-Actin (A5316, Sigma), and mouse anti-Raf-B (sc-5284, Santa 847

Cruz Biotechnology). For immunofluorescence staining, cells were fixed and permeabilized 848

followed by staining with anti-H3K9Me3 antibody (07-442 from Millipore, Billerica, MA) or anti-849

HP1antibody (ab56978 from Abcam). Cells were then labeled with Alexa Fluor488 donkey 850

anti-Rabbit secondary Ab or Alexa Fluor568 goat anti-Rabbit secondary antibody (Invitrogen, 851

Carlsbad, CA) before treated with anti-fade gold mounting medium with DAPI (Invitrogen, 852

Carlsbad, CA). Images were examined and processed by a Zeiss LSM 700 confocal 853

microscope with total 630X magnification. Representative images are shown. To measure 854

H3K9Me3 level by western blotting, total histone was extracted using HCL following a standard 855

protocol (Abcam) and detected with rabbit anti-Histone H3 (61277, Active Motif) or rabbit anti-856

Histone H3K9Me3 (ab8898, Abcam) antibody; ratios of H3K9Me3 to total H3 did not correlate 857

with H3K9Me3 focus formation (Supplementary Fig. 26). 858

Colony formation assays. Anchorage-independent growth was assayed with 859

p’mel/BRAFV600E cells (provided by Dr. Hans Widlund, Dana-Farber Cancer Institute). Briefly, 860

2,500 cells were mixed in HAMF12 plus 10% FBS medium containing 0.4% SeaKem LE 861

agarose from Lonza, and plated on top of the bottom layer with 0.65% agarose prepared in 862

same HAMF12 medium in 6-well plate. After the agarose is solidified, each well was covered 863

with 0.5 mL feeding medium which was refreshed twice a week. Colonies were counted and 864

imaged between 3-4 weeks after seeding under a regular microscope with total magnification of 865

40X. Colonies were counted from three wells of each condition, and significant difference from 866

the control was assessed by t-test assuming unequal variance. 867

Chromatin immunoprecipitation of PARP1, H3K4me3, and RNA PolII. Primary human 868

melanocytes (HEMn-LP, Invitrogen) were fixed with 0.1% formaldehyde when 80-90% 869

confluent. About 1 x 107 cells were then sheared by sonication using a Bioruptor (Diagenode) at 870

high setting for 10 cycles with 20 sec on and 30 sec off. Chromatin were then purified and 871

immunoprecipitated following the instructions of Active Motif CHIP-IT high sensitivity kit. 5-10 µg 872

sheared chromatin were used for each immunoprecipitation with antibodies against 873

PARP1,46D11 (#9532, Cell Signaling Technology), H3K4me3 (ab8580, Abcam), RNA PolII 874

CTD (17-672, Millipore), RNA PolII CTD Phospho-Ser5 (ab5408, abcam) or normal rabbit IgGs 875

(ab46540, Abcam) and normal mouse IgGs (17-672, Millipore). Purified pulled-down DNA was 876

assayed by SYBR Green qPCR for enrichment of target sites across MITF-M promoter using 877

primers listed in Supplementary Table 14. 878

Statistical analyses 879

All cell-based experiments were repeated at least three times with separate cell cultures, except 880

for Fig3a (repeated twice), Fig4d-e (six technical replicates), and Fig5d-e (repeated twice). 881

When a representative set was shown, replicate experiments displayed similar patterns. For all 882

the plots, individual data points are shown with median or mean, range (maximum and 883

minimum), and 25 & 75 percentile (where applicable). Statistical method, number of data points, 884

and number and type of replicates are indicated in each figure legend. 885

Accession codes 886

dbGAP Accession: phs000178.v9.p8 887

dbGaP Accession: phs000424.v6.p1 888

GEO Accession: GSE60666 889

890

DATA AVAILABILITY STATEMENT 891

The data generated during the current study have been deposited in NCBI's Gene Expression 892

Omnibus (Edgar et al., 2002) with the accession code GSE99221 (super series) which includes 893

GSE99193 (genotype data) and GSE78995 (expression data). 894

REFERENCES 895

7. Law, M.H. et al. Genome-wide meta-analysis identifies five new susceptibility loci for cutaneous 896

malignant melanoma. Nat Genet 47, 987-95 (2015). 897

26. Consortium, E.P. An integrated encyclopedia of DNA elements in the human genome. Nature 898

489, 57-74 (2012). 899

28. Boersema, P.J., Raijmakers, R., Lemeer, S., Mohammed, S. & Heck, A.J. Multiplex peptide stable 900

isotope dimethyl labeling for quantitative proteomics. Nat Protoc 4, 484-94 (2009). 901

29. Hubner, N.C., Nguyen, L.N., Hornig, N.C. & Stunnenberg, H.G. A quantitative proteomics tool to 902

identify DNA-protein interactions in primary cells or blood. J Proteome Res 14, 1315-29 (2015). 903

30. Butter, F. et al. Proteome-wide analysis of disease-associated SNPs that show allele-specific 904

transcription factor binding. PLoS Genet 8, e1002982 (2012). 905

39. Rolli, V., O'Farrell, M., Menissier-de Murcia, J. & de Murcia, G. Random mutagenesis of the 906

poly(ADP-ribose) polymerase catalytic domain reveals amino acids involved in polymer 907

branching. Biochemistry 36, 12147-54 (1997). 908

58. Howie, B.N., Donnelly, P. & Marchini, J. A flexible and accurate genotype imputation method for 909

the next generation of genome-wide association studies. PLoS Genet 5, e1000529 (2009). 910

59. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol 9, R137 (2008). 911

60. Dignam, J.D., Lebovitz, R.M. & Roeder, R.G. Accurate transcription initiation by RNA polymerase 912

II in a soluble extract from isolated mammalian nuclei. Nucleic Acids Res 11, 1475-89 (1983). 913

61. Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-914

range mass accuracies and proteome-wide protein quantification. Nat Biotechnol 26, 1367-72 915

(2008). 916

917